Survey banner
The Dataiku Community is moving to a new home! Some short term disruption starting next week: LEARN MORE

No model-specific details available - Keras/TensorFlow

cooxky
Level 3
No model-specific details available - Keras/TensorFlow

Dear All,

I used Dataiku deep learning interface to build a Keras/Tensorflow model using functional API.

Training works just fine, but there are no details about the model itself, instead I can see "No model-specific details available" (screen attached).

Is that the way it's supposed to work or am I missing something?

I wouldn't mind seeing some more details there - architecture of the model, learning rate etc.

 

 

Operating system used: Amazon Linux


Operating system used: Amazon Linux

0 Kudos
7 Replies
AlexT
Dataiker

Hi,
"No model-specific details available" can appear when the model is not probabilistic
If you did not implement the method  predict_proba in their estimator class.

Since DSS is not able to compute anything that requires probs (i.e. only make sense if you have problems available),

If this is indeed the case you can see that in the logs with the line:
Cannot draw background rows and compute column importance: model is not probabilistic

Thanks

0 Kudos
cooxky
Level 3
Author

Hello AlexT,

Thanks for your asnwer.

 

As I mentioned at the begging of my post, I used Dataiku interface for deep learning.

Which means I build a model using TensorFlow functional API.

Dataiku documentation doesn't tell you to implement predict_proba() method when you're using that, as you do when you create a class deriving from sklearn BaseEstimator.

I don't think that's the case with TensorFlow, right?

 

I did a test, I used default code that Dataiku puts in the "Architecture" tab of deep learning GUI, which is:

 

def build_model(input_shapes, n_classes=None):

    # This input will receive all the preprocessed features
    # sent to 'main'
    input_main = Input(shape=input_shapes["main"], name="main")

    x = Dense(64, activation='relu')(input_main)
    x = Dense(64, activation='relu')(x)

    predictions = Dense(n_classes, activation='softmax')(x)

    # The 'inputs' parameter of your model must contain the
    # full list of inputs used in the architecture
    model = Model(inputs=[input_main], outputs=predictions)

    return model

 

As you can see, final layer uses softmax, which returns probabilities per class.

Still, the only thing I can see in the GUI is "No model-specific details available".

I also checked the logs and there is no "Cannot draw background rows and compute column importance: model is not probabilistic" there.

 

0 Kudos
AlexT
Dataiker

Hi @cooxky ,

Could you please open a support ticket and attach the model training diagnostics or training log if you are an older DSS version? From the visual analysis page

Screenshot 2023-11-15 at 7.37.27 AM.png

0 Kudos
cooxky
Level 3
Author

Thanks AlexT, I've opened a support ticket: 58374

0 Kudos
cooxky
Level 3
Author

@AlexT I received info that model info is only available for time series models.

However, on your print screen I can see some info available.

Could you please share on how you did it?

The only way I can think of is wrapping your deep learning model in sklearn BaseEstimator and then possibly turning it to a plugin.

0 Kudos
AlexT
Dataiker

Yes, that's correct this is only implemented for Time Series deep learning models.
And regular prediction models e.g sklearn models. What info exactly are you hoping to obtain under "model-specific details"?

Kind Regards,

0 Kudos
cooxky
Level 3
Author

Well, things like info about learning rate, optimizer/initializer used, some basic info about architecture like nr of hidden layers.

0 Kudos